--- title: "LangChain / LangGraph - 工具配置与选型 - AI·外脑 | 思行致远" canonical: "https://siyuanailab.com/knowledge/toolkits/langchain-langgraph/" dateModified: "2026-07-08" datePublished: "2026-07-08" --- 打开 AI·外脑目录 AI·外脑 [返回专栏首页](https://siyuanailab.com/knowledge/) × 站内搜索 ⌕ 主要栏目 - [AI 核心概念](https://siyuanailab.com/knowledge/ai-core/) - [工作流](https://siyuanailab.com/knowledge/workflows/) - [工具配置与选型](https://siyuanailab.com/knowledge/toolkits/) - [信源与标准](https://siyuanailab.com/knowledge/sources/) - [场景案例 待开发](https://siyuanailab.com/knowledge/scenarios/) - [职业与能力](https://siyuanailab.com/knowledge/career/) 专题 - [从这里开始](https://siyuanailab.com/knowledge/start/) - [获取与安装](https://siyuanailab.com/knowledge/install/) - [建设日志](https://siyuanailab.com/knowledge/ops/) [工具配置与选型子目录](https://siyuanailab.com/knowledge/toolkits/) llm-image（8） - [Adobe Firefly](https://siyuanailab.com/knowledge/toolkits/adobe-firefly/) - [通义万相](https://siyuanailab.com/knowledge/toolkits/aliyun-wanxiang-image/) - [字节即梦 AI](https://siyuanailab.com/knowledge/toolkits/bytedance-jimeng/) - [可灵 AI](https://siyuanailab.com/knowledge/toolkits/kling-ai-image/) - [Midjourney](https://siyuanailab.com/knowledge/toolkits/midjourney/) - [OpenAI GPT Image](https://siyuanailab.com/knowledge/toolkits/openai-gpt-image/) - [FLUX](https://siyuanailab.com/knowledge/toolkits/flux/) - [Stable Diffusion](https://siyuanailab.com/knowledge/toolkits/stable-diffusion/) llm-china（6） - [阿里通义千问 Qwen](https://siyuanailab.com/knowledge/toolkits/aliyun-qwen/) - [百度文心一言](https://siyuanailab.com/knowledge/toolkits/baidu-ernie/) - [字节豆包](https://siyuanailab.com/knowledge/toolkits/bytedance-doubao/) - [DeepSeek](https://siyuanailab.com/knowledge/toolkits/deepseek/) - [月之暗面 Kimi](https://siyuanailab.com/knowledge/toolkits/moonshot-kimi/) - [智谱清言](https://siyuanailab.com/knowledge/toolkits/zhipu-chatglm/) llm-video（9） - [通义万相视频](https://siyuanailab.com/knowledge/toolkits/aliyun-wanxiang-video/) - [字节即梦 AI 视频](https://siyuanailab.com/knowledge/toolkits/bytedance-jimeng-video/) - [Hailuo AI（海螺 AI）](https://siyuanailab.com/knowledge/toolkits/hailuo-ai/) - [可灵 AI 视频](https://siyuanailab.com/knowledge/toolkits/kling-ai-video/) - [Pika](https://siyuanailab.com/knowledge/toolkits/pika/) - [智谱清影](https://siyuanailab.com/knowledge/toolkits/zhipu-qingying/) - [Google Veo](https://siyuanailab.com/knowledge/toolkits/google-veo/) - [Luma Dream Machine](https://siyuanailab.com/knowledge/toolkits/luma-dream-machine/) - [Runway](https://siyuanailab.com/knowledge/toolkits/runway/) llm-international（4） - [Anthropic Claude](https://siyuanailab.com/knowledge/toolkits/anthropic-claude/) - [Google Gemini](https://siyuanailab.com/knowledge/toolkits/google-gemini/) - [OpenAI / ChatGPT](https://siyuanailab.com/knowledge/toolkits/openai-chatgpt/) - [Meta Llama](https://siyuanailab.com/knowledge/toolkits/meta-llama/) hardware-health（2） - [Apple Watch Series 11](https://siyuanailab.com/knowledge/toolkits/apple-watch-series-11/) - [Garmin Venu 4](https://siyuanailab.com/knowledge/toolkits/garmin-venu-4/) productivity-image（1） - [Canva](https://siyuanailab.com/knowledge/toolkits/canva/) productivity-video（1） - [CapCut](https://siyuanailab.com/knowledge/toolkits/capcut/) agent-china（3） - [Coze / Coze Studio](https://siyuanailab.com/knowledge/toolkits/coze/) - [Dify](https://siyuanailab.com/knowledge/toolkits/dify/) - [FastGPT](https://siyuanailab.com/knowledge/toolkits/fastgpt/) hardware-audio（3） - [DJI Mic 3](https://siyuanailab.com/knowledge/toolkits/dji-mic-3/) - [DJI Mic Mini 2](https://siyuanailab.com/knowledge/toolkits/dji-mic-mini-2/) - [Zoom H1essential](https://siyuanailab.com/knowledge/toolkits/zoom-h1essential/) productivity-audio（1） - [ElevenLabs](https://siyuanailab.com/knowledge/toolkits/elevenlabs/) knowledge-base-cloud（4） - [飞书知识库](https://siyuanailab.com/knowledge/toolkits/feishu-wiki/) - [Notion](https://siyuanailab.com/knowledge/toolkits/notion/) - [腾讯 ima.copilot](https://siyuanailab.com/knowledge/toolkits/tencent-ima/) - [Atlassian Confluence](https://siyuanailab.com/knowledge/toolkits/atlassian-confluence/) productivity-presentation（1） - [Gamma](https://siyuanailab.com/knowledge/toolkits/gamma/) agent-coding（4） - [GitHub Copilot](https://siyuanailab.com/knowledge/toolkits/github-copilot/) - [Claude Code](https://siyuanailab.com/knowledge/toolkits/claude-code/) - [Cursor](https://siyuanailab.com/knowledge/toolkits/cursor/) - [OpenAI Codex](https://siyuanailab.com/knowledge/toolkits/openai-codex/) knowledge-base-local（2） - [Logseq](https://siyuanailab.com/knowledge/toolkits/logseq/) - [Obsidian](https://siyuanailab.com/knowledge/toolkits/obsidian/) ai-search-chinese（3） - [秘塔 AI 搜索](https://siyuanailab.com/knowledge/toolkits/metaso/) - [夸克 AI 搜索](https://siyuanailab.com/knowledge/toolkits/quark-ai/) - [天工 AI](https://siyuanailab.com/knowledge/toolkits/tiangong-ai/) decision（4） - [从想试试到用起来：3 步最小路径](https://siyuanailab.com/knowledge/toolkits/minimal-path/) - [角色优先：我是谁，我该用什么](https://siyuanailab.com/knowledge/toolkits/role-first/) - [用例优先：我该先解决什么问题](https://siyuanailab.com/knowledge/toolkits/use-case-first/) - [工具评估 8+4 维](https://siyuanailab.com/knowledge/toolkits/evaluation-framework/) ai-search-general（1） - [Perplexity](https://siyuanailab.com/knowledge/toolkits/perplexity/) agent-open-source（2） - [CrewAI](https://siyuanailab.com/knowledge/toolkits/crewai/) - [LangChain / LangGraph](https://siyuanailab.com/knowledge/toolkits/langchain-langgraph/) hardware-wearable（3） - [Ray-Ban Meta AI Glasses](https://siyuanailab.com/knowledge/toolkits/ray-ban-meta-glasses/) - [XREAL One Pro](https://siyuanailab.com/knowledge/toolkits/xreal-one-pro/) - [Apple Vision Pro](https://siyuanailab.com/knowledge/toolkits/apple-vision-pro/) 工具配置与选型 · 进阶 # LangChain / LangGraph 杨思远 / 思行致远 · 2026/7/8 · 置信度：高 Agent 框架开源PythonJavaScriptRAG Agent 和 LLM 应用工程的事实标准之一，生态集成广，适合开发者搭建可观测、可部署的 Agent 应用。 # LangChain / LangGraph > Agent 和 LLM 应用工程的事实标准之一，生态集成广，适合开发者搭建可观测、可部署的 Agent 应用。 --- ## 一、是什么 LangChain 是一个开源 Python/JS 框架，帮助开发者把大语言模型、数据源、工具链组合成可运行的应用。LangGraph 是其生态中的工作流编排层，适合构建多步骤 Agent 和状态机。 ## 二、为什么重要 如果你想把 LLM 从“玩具”变成“产品”，LangChain 提供了模型接入、提示词管理、记忆、工具调用、RAG 等通用抽象，是最广泛使用的 LLM 应用框架之一。 ## 三、核心机制 / 关键信息 - **官方名称**：LangChain / LangGraph - **官网**：[https://www.langchain.com](https://www.langchain.com/) - **GitHub**：[https://github.com/langchain-ai/langchain](https://github.com/langchain-ai/langchain) - **RSS / 动态**：[https://blog.langchain.com](https://blog.langchain.com/) - **当前主要版本 / 模型矩阵**：langchain 1.x；LangGraph / LangSmith 配套 - **国内可用性**：开源包可用；LangSmith 云服务需海外网络 - **定价模式**：开源框架免费；LangSmith / 云服务 Freemium + 企业定价 ## 四、典型用法（0→1→2→3） - **Step 0：安装并跑通第一个 Chain**：连接一个 LLM，完成一次问答。 - **Step 1：加入 Retriever**：用向量数据库做 RAG，让模型回答私有资料。 - **Step 2：用 LangGraph 编排多步 Agent**：让模型决定调用哪些工具。 - **Step 3：接入 LangSmith 观测**：追踪调用链、成本和错误。 ## 五、常见误区与替代方案 - **误区**：非技术用户必须学 LangChain。普通用户先用 Dify/Coze 等低代码平台即可。 - **误区**：LangChain 是模型。它是一个框架，本身不提供模型。 - **替代方案**：需要更简单可视化编排时对比 Dify；需要多 Agent 角色协作时对比 CrewAI。 ## 六、延伸阅读 / 相关卡片 - [[Dify]] - [[CrewAI]] - [[GitHub Copilot]] - [LangChain Docs](https://python.langchain.com/docs/introduction/) --- ## 使用提示（AI 引用块） - **适用场景**：LLM 应用开发、RAG、Agent 工作流、需要高度自定义的工程团队。 - **不适用场景**：非技术用户、没有开发能力、只需要简单聊天机器人。 - **常见错误**：为了用框架而用框架，把简单任务复杂化。 --- *这张卡片由杨思远基于真实工作流和实测经验整理。如发现信息过时或有误，可通过 /contact 反馈。* ### 适用场景与行动 **场景：**需要开发可扩展的 LLM 应用、Agent 工作流或 RAG 系统 **适合谁：**开发者、技术合伙人、有工程能力的小团队 **下一步：**用 LangChain 写一个最简单的链式调用示例，连接 OpenAI API 完成一次问答。 ### 相关卡片 - CrewAI - Dify - Claude Code - Cursor 这张卡片由杨思远基于真实工作流和实测经验整理。 如发现信息过时或有误，可通过 [/contact](https://siyuanailab.com/contact) 反馈。 [← 上一张：Google Veo](https://siyuanailab.com/knowledge/toolkits/google-veo/) [下一张：Luma Dream Machine →](https://siyuanailab.com/knowledge/toolkits/luma-dream-machine/) [返回 AI·外脑入口](https://siyuanailab.com/knowledge/)